FinancialInstitutionsCenter

Lessons from Behavioral Finance forRetirement Plan Design

byOlivia MitchellStephen Utkus

03-34

The Wharton Financial Institutions Center

The Wharton Financial Institutions Center provides a multi-disciplinary research approach tothe problems and opportunities facing the financial services industry in its search forcompetitive excellence. The Center's research focuses on the issues related to managing riskat the firm level as well as ways to improve productivity and performance.

The Center fosters the development of a community of faculty, visiting scholars and Ph.D.candidates whose research interests complement and support the mission of the Center. TheCenter works closely with industry executives and practitioners to ensure that its research isinformed by the operating realities and competitive demands facing industry participants asthey pursue competitive excellence.

Copies of the working papers summarized here are available from the Center. If you wouldlike to learn more about the Center or become a member of our research community, pleaselet us know of your interest.

Franklin Allen Richard J. HerringCo-Director Co-Director

The Working Paper Series is made possible by a generousgrant from the Alfred P. Sloan Foundation

Lessons from Behavioral Finance for Retirement Plan Design

Olivia Mitchell and Stephen Utkus

PRC WP 2003-6

Pension Research Council Working Paper

Pension Research Council

The Wharton School, University of Pennsylvania 3641 Locust Walk, 304 CPC Philadelphia, PA 19104-6218

Tel: 215.898.7620 • Fax: 215.898.0310 • Email: prc@wharton.upenn.edu http://prc.wharton.upenn.edu/prc/prc.html

Pension Research Council Working Papers are intended to make research findings available to other researchers in preliminary form, to encourage discussion and suggestions for revision before final publication. Opinions are solely those of the authors. This paper is to appear in Money Matters: Shaping Retirement Decisionmaking (forthcoming). Edited by Olivia S. Mitchell and Stephen P. Utkus. Oxford: Oxford University Press. ©2003 Pension Research Council of the Wharton School of the University of Pennsylvania. All Rights Reserved.

Lessons from Behavioral Finance for Retirement Plan Design

Olivia S. Mitchell and Stephen P. Utkus

Abstract

This paper evaluates some of the key lessons of behavioral economics and finance research over the last decade for pension plan design. We divide the discussion into the natural phases of the retirement saving life cycle: accumulation, investment, and decumulation. After reviewing the lessons of behavioral finance, we conclude by outlining plan design alternatives that would be of use to plan sponsors and policymakers seeking to design more cost-effective and efficient retirement plans for the future. Olivia S. Mitchell Stephen P. Utkus Professor of Insurance and Risk Management Principal The Wharton School Vanguard Center for Retirement Research 3641 Locust Walk, Rm. 304 CPC 100 Vanguard Boulevard, J24 Philadelphia PA, 19104-6218 Malvern, PA 19355 T: 215/746-5706 T: 610/669-6308 Email: mitchelo@wharton.upenn.edu Email: steve_utkus@vanguard.com

Lessons from Behavioral Finance

for Retirement Plan Design

Olivia S. Mitchell and Stephen P. Utkus

Participant-directed defined contribution (DC) plans have become the cornerstone of the private-

sector retirement system around the world. In the U.S., participant choice has spread not only to

pensions, but to a great many other aspects of the employee benefit package as well including health care

plans, flexible benefit programs, and time-off arrangements. The trend toward giving participants more

choice also underlies recent proposals to reform Social Security by adding personal accounts, and

Medicare proposals to permit seniors to choose whether they want a public versus a privately-managed

healthcare plan. Participant-managed DC plans are the main feature of national pension reforms already

implemented in many Latin American nations, as well as in Germany, Sweden, and most recently, Russia.

Underlying this global movement spurring participant choice is an implicit assumption about

behavior: that the employee-citizen to whom the responsibility of choice has been handed is a well-

informed economic agent who acts rationally to maximize his self interest. To this end, it is assumed that

he can interpret and weigh information presented regarding options offered by employers and

governments, appropriately evaluate and balance these choices, and then make an informed decision

based on a weighing of the alternatives.

Recently, however, a different perspective has emerged regarding how “real” people make

economic decisions, one developed by social scientists working at the interface of economics, finance,

psychology, and even sociology. This perspective is consistent with the fundamental economic

proposition that people can and do try to maximize their self-interest, but it also recognizes that such

decisions are often made with less-than perfect outcomes. In the real world, peoples’ decisions are

subject to “bounded rationality,” as the Herbert Simon called it (Simon 1955). Certain types of decisions

and problems may be simply too complex for individuals to master on their own. There is also what

2

Mullainathan and Thaler (2000) call “bounded self-control”—individuals have the right intentions or

beliefs, but they lack the willpower to carry out the appropriate changes in behavior. And last, there is the

problem of “bounded self-interest” or “bounded selfishness” (Mullainathan and Thaler, 2000). This

acknowledges that many people do seek to maximize their personal welfare, yet they prove far more

cooperative and altruistic than economic theory predicts they will be.

These new notions of how people make decisions have spurred the rapidly growing fields of behavioral

economics and finance. The central question addressed by this research is how markets work and how

consumers make decisions when some (or even many) people labor under such mental or emotional

constraints and complications.1 This research is having a profound impact on the way analysts now view

varied aspects of economic and financial life, including the ways in which we understand how people

decide to save, invest, and consume.2 The goal of the present chapter is to evaluate key aspects of this

new behavioral research in light of what it tells us about better ways to design and manage retirement

systems. In what follows, therefore, we analyze what insights this literature offers us on how workers

decide to save, how they manage their retirement investments, and ultimately how they decide to draw

down their assets in retirement. Our aim, in particular, is to understand how workers and retirees might

deviate from the rational, all-knowing economic agents that underpin economic theory and often

retirement plan design. Finally, we discuss implications of this literature for plan sponsors and

policymakers who must design, regulate, and evaluate the institutions that help provide for economic

security in old age.

The Decision to Save

Understanding why people save, and what they invest in, are questions of central important to

economists and policymakers. With the growth of defined contribution saving plans in the US and around

the world, especially plans having a 401(k) or employee contributory feature, it is clear that having a

meaningful retirement benefit depends increasingly on participants’ decisions to save and invest in their

retirement plans.

3

Neoclassical economic theory casts the saving outcome as the result of people trading off current

versus future consumption. Thus households are thought to compare the benefit gained from consuming

their income today, with the benefits of deferring some of that income into the future. This is what is

thought to drive contributions to a 401(k) or individual retirement account, with the goal being to save for

retirement. The life-cycle model of saving posits that individuals are rational planners of their

consumption and saving needs over their lifetimes, taking into account the interests of their heirs

(Modigliani and Brumberg, 1954). During their younger years, workers tend to be net dissavers,

borrowing from the future by means of debt to boost current consumption; middle-aged individuals

become net savers and purchasers of financial assets and enter “accumulation” phase, during which they

stockpile assets for the final, retired phase of life. As labor earnings decline or disappear, people then

decumulate or draw down their financial assets to finance old-age consumption. According to the tenets

of life-cycle theory, people will logically develop assets for retirement that will be sufficient to protect

them from unexpected declines in their standard of living in old age.

On balance, the life-cycle theory is thought to do a reasonable job of explaining patterns of

household saving behavior. Saving generally rises with income and age, and it is positively associated

with education and total wealth. Young households generally have more debt than assets, while prime-

aged households do appear to begin saving more and accumulating financial holdings. Finally, in

retirement, people do tend to consume portions of their financial assets as they age.3

On the other hand, some saving behavior appears to be at odds with the theory. Consider, first, a

fundamental question: how good are households at calculating an appropriate saving goal for retirement?

Arguably, if the lifecycle analysis is true, households should have some demonstrated skill at estimating

their needs for retirement, and analysis of actual savings behavior should demonstrate some reasonably

widespread competency at the task. Yet superficially the retirement savings problem is perhaps an ideal

illustration of Simon’s “bounded rationality”. Being good at retirement savings requires accurate

estimates of uncertain future processes including lifetime earnings, asset returns, tax rates, family and

health status, and longevity. In order to solve this problem, the human brain as a calculating machine

4

would need to have the capacity to solve many decades-long time value of money problems, with massive

uncertainties as to stochastic cash flows and their timing.

In fact, survey and empirical research suggests that individuals are not particularly good at the

retirement savings problem. Relatively few people feel they are able to plan effectively for retirement

(Lusardi, this volume). Indeed, surveys repeatedly find that fewer than 40 percent of US workers have

calculated how much they will need to retire on, 30 percent have not saved anything for retirement, and

only 20 percent feel very confident about having enough money to live comfortably in retirement (EBRI

2003).

Furthermore, the empirical evidence suggests that failing to save enough also has serious negative

consequences. A recent study of post-retirement consumption patterns indicates that US workers

experience an unexpected decline in their standard of living after retirement (Bernheim et al., 2001). This

consumption drop is even more precipitous in the UK (Banks et al., 1998). 4 Other research suggests that

only 30 percent of pre-retirees are fully prepared for retirement at age 65 in the US (Moore and Mitchell,

2000). Of the remaining group, another 30 percent is likely to close the savings gap by age 65, though

this hardly appears to be evidence of a long-term, life-long rational planner at the heart of the life-cycle

model. Finally, fully 40% appear unlikely to achieve a reasonable standard of replacement income by age

65. The numbers are much more pessimistic if retirement is planned for age 62, when the median

American typically retires.

Behavioral economists would not find it surprising that people struggle with retirement saving in

view of the problem’s complexity. Indeed, many would take it as prima facie evidence that large groups

of workers do not “get the saving problem right,” contrary to the assumption of rationality and wise

planning underlying the life-cycle model.

The Problem of Self-control. What might explain this lack of retirement preparation? Behavioralists

tend to rely on a straightforward psychological explanation called “lack of willpower.” This explanation

is often described as “bounded self-control”: that is, people try to save for retirement, but they too often

prove to be limited in their capacity or desire to execute intentions (Thaler and Shefrin, 1981). In a sense,

5

saving for retirement requires behavior similar to those undertaken in other behavior modification

programs such as exercising, dieting, quitting smoking, or following through on New Year’s resolutions.

It would seem that while people intellectually “understand” the benefits of a specific behavior, and they

may even have some idea of how to get started, they have difficulty implementing their intentions. Too

often, they struggle to take action, and when they do act, their behaviors are often half-hearted or

ineffective.

What evidence is there that problems of self-control may be important deterrents to saving for

retirement? One body of researchers offers practical and theoretical insights into how individuals make

tradeoffs regarding risk and time. Psychologists have shown that peoples’ near-term discount rates are

much higher than their long-term discount rates (Laibson et al., 1998). In Thaler’s (1981) formulation,

people confronting long-term decisions can exhibit high levels of patience. For instance, they might say

“If I can receive an apple in 100 days and two apples in 101 days, I’ll be happy to wait the extra day for

another apple.” But when the decision shifts to the present, their patience wears thin and they think: “I’d

rather have an apple today than wait for two tomorrow.”5

In standard time value of money calculations, discount rates are postulated to remain constant

over time, so they do not vary today, tomorrow, or a year from now. Given this assumption, one dollar

saved today would be perceived to be worth exponentially more (e.g. $5.74) in 30 years’ time (Figure 1).

But when individuals are “hyperbolic discounters,” they apply high discount rates to the near term and

lower discount rates to the future. In this case, one dollar’s worth of saving today is perceived as growing

more rapidly in the short run, and less in the longer run. Hence the incremental gains from extending

one’s time horizon are perceived to be relatively modest, compared to those of the conventional

“exponential discounter.” As illustrated by the increasing slope of the exponential line, exponential

savers foresee ever-increasing rewards to deferring consumption by another year. For hyperbolic

discounters, rewards are left to accelerate quickly, and then taper off. Put simply, workers who are

hyperbolic discounters place a lower value on future benefits and overvalue the present. The application

6

to retirement is clear: they will overconsume today and undersave, as a result of self-control problems

when it comes to saving for retirement.

Figure1 here

Decision theorists working in this vein seek to understand the self-control problem in a deeper

way, delving into the structure and processing mechanisms of the human brain. For instance, Weber (this

volume) notes that the brain consists of an older, limbic system shared with lower-order animals, which is

the source of emotional or affective decision-making; and a more “modern” cerebral layer, which is a

late-stage evolutionary trait in humans and the source of conceptual or symbolic processing. Perhaps

because of this, processing of emotions typically involves gauging risk in terms of two components:

“dread risk,” or the potential for catastrophe, and “uncertainty risk,” involving a generalized fear of the

unknown or the new. Weber suggests that retirement risks rate low along both dimensions: few people

have a palpable fear of impending disaster or of great uncertainty in their retirement planning, as

compared to other risks in their lives. In Weber’s framework, the self-control problem of retirement

saving must join both cerebral and emotional decision-making simultaneously, if people are to be

prompted to take effective action. For example, if one were to experience the risks of retirement in the

present so as to stimulate the brain’s affective system, people might attempt a real-world experiment such

as attempting to live on, say, two-thirds of their income for the next month.

Whether viewed from an economics or a decision theoretic perspective, the self-control problem

supports the view of a wide divergence between individuals’ desires and their actual behaviors

(Saliterman and Sheckley, this volume). A survey of 10,000 employees at a single firm found that 68

percent of participants said their retirement savings rate was “too low” (Choi et. al., 2001a). When

queried, they reported that they should be saving 14 percent of average earnings, whereas in fact, they

were only saving about 6 percent. (The remaining one-third of the participants believed their saving

patterns were just about right and fewer than 1 percent felt they were saving too much.) Similarly, in

Clark et al. (this volume), retirement plan participants reported that they knew they were saving less than

they should. In other words, a key obstacle to saving more is not necessarily lack of awareness, but rather

7

the ability to take action on the knowledge. The difficult task is to overcome hyperbolic discounting, to

merge conceptual and affective reasoning into a course of effective action.

In recognition of such problems, people often seek to protect themselves through the use of

commitment devices, or mechanisms that help foster desirable changes in behavior (cf Laibson 1997, and

Laibson et al. 1998). Commitment devices for saving may be an analogue of the fad diet: one way of

imposing some degree of discipline on one’s wayward behavior is to create some seemingly arbitrary

rules about what one can and cannot eat. “Pay yourself first” is a standard commitment device used by

financial planners seeking to encourage disciplined saving and budgeting; it is also the principle

underlying US payroll-deduction 401(k) plans. These plans are one of the most successful commitment

devices in current use, and they are formulated such that contributions are automatically deducted from

workers’ pay before the money can be spent. Participation rates in 401(k)-type plans, where payroll

deduction is the norm, are at least four times as high as for Individual Retirement Accounts, where

structured payroll deductions are uncommon; according to 1997 tax return data, some 27 percent of

workers contributed to workplace savings plans, compared with 6 percent contributing to an IRA (CBO,

2003). Other commitment devices include tax refunds and Holiday Clubs, where individuals engage in

seemingly irrational economic activity (e.g. loaning money to the government or to their local banks at

below-market rates) in exchange for discipline at accumulating savings. Withdrawal restrictions on IRAs

and 401(k)s and other retirement plans also appear to be commitment devices: once the money is

allocated to these plans, a psychological and financial hurdle is imposed on accessing the money, helping

to counteract lapses in personal willpower.

Other evidence that individuals vary in their capacity for self-control and financial discipline

comes from industry surveys of workers’ savings and planning behavior. Ameriks, Caplin and Leahy

(2002) as well as Lusardi (this volume) find that workers’ “propensity to plan” has strong positive

influence on retirement wealth accumulation. MacFarland, Marconi and Utkus (this volume) indicate that

as many as half of pension participants are dis- or uninterested in the financial and retirement planning

activities thought necessary to plan successful retirement. In fact, a “planner” paradigm, where the

8

individual consciously pursues retirement saving and investment goals in a disciplined, systematic way,

appears to apply to only about half of the retirement plan population. The other half appears singularly

unable to impose the self-control needed to solve this problem.

Framing and Default Choices. Many individuals deviate from standard economic theory in another

important way: they can be easily influenced by decision framing. Rational economic agents would not

be expected to vary their responses to a question based on how it is asked. But in practice, many people

do exactly that, both in the savings area and, as we show note later, in investment decision-making as

well. A by-now classic example of decision framing arises with automatic enrollment in retirement

saving plans. Under the traditional (non-automatic) approach, the employee would have to make a

“positive election” to join the 401(k) plan. By contrast, with automatic enrollment, the employee would

be signed up by the employer for the plan at a given percentage contribution rate, and the employee

retains the right to opt out of this decision.

This simple rephrasing of the saving question elicits a dramatically different response in plan

participation rates. Madrian and Shea (2001) have powerfully shown that when workers are required to

opt-in, the default decision (or the non-decision) is to save nothing; by dramatic contrast, with automatic

enrollment, the default decision proves to be that people save at the rate specified by the employer. For

one large US firm, plan participation rates jumped from 37 percent to 86 percent for new hires after

automatic enrollment was introduced (Figure 2). What this suggests, in the end, is that many workers do

not have particularly firm convictions about their desired savings behavior. Merely by rephrasing the

question, their preferences can be changed—from not saving to saving.

Figure 2 here

The impact of automatic enrollment is not just an illustration of framing questions but also part of

a broader behavioral phenomenon, namely the power of the “default option” and its influence on

decision-making. When confronted with difficult decisions, individuals tend to adopt heuristics

(shortcuts) that simplify the complex problems they face. One simple heuristic is to accept the available

9

default option—i.e., rather than making an active choice, accept the choice made by others. And, as

noted in 401(k) enrollment, the simplest default is the non-decision: do nothing.

An emerging literature indicates that individual behavior is easily swayed by default choices.6

Again, automatic enrollment provides another illustration of the unexpected effects of default behavior. It

turns out that while automatic enrollment boosts the number of individuals saving in a retirement plan, it

might not actually increase total plan savings (Choi et al. 2001b). The reason is that, when automatically

enrolled, people who would have voluntarily enrolled in the plan at higher contribution rates or chosen

more aggressive investments decide to stick with the low saving rate and conservative investment option

set by their employer. Thus the positive effect is that saving rises for people who formerly did not

participate, but an unexpectedly negative result is that saving falls for those who would have enrolled at

higher rates and in more aggressive options, but instead elected to adopt the employer’s defaults. On net,

it appears that these two effects can largely offset one another.

More broadly, Choi et al. (2003) develop a model of a procrastination- and default-driven saver

and the choice of optimal savings rates. That study argues that optimal defaults for such savers are, in

effect, the corner points or defaults of the plan savings problem—a saving rate of 0 percent, a saving rate

equal to the employer matching contribution, and a saving rate at the maximum allowed by the plan.

Both their theoretical models and the practical evidence on automatic enrollment underscore how

profound the impact can be of the selection of a default option.

Inertia and Procrastination. Evidence on automatic enrollment has also revealed another anomaly

about individuals and their saving behavior: the important impact that inertia or procrastination plays on

decision-making. In Madrian and Shea’s (2001) analysis of automatic enrollment, they showed that the

benefit of higher plan participation rates appeared to be offset by a profound level of inertia. Most

participants remained at the default savings and conservative investment choices set for them by their

employer. Once enrolled, participants made few active changes to the contribution rates or investment

mixes selected for them by their employer; rather, they simply stayed with what was assigned to them.

10

Another analysis (Choi et al., 2001a), explored how inertia and default behavior influenced

several other defined contribution plan activities: enrollment, cash distributions at termination, the match

level, eligibility, and the impact of education, among others. The authors concluded that, more often than

not, many participants followed “the path of least resistance” in their decision-making—in effect, making

the easiest, rather than the best, decision. Again, the persistence of inertia and what might be called a

passive approach to decision-making are both indicative of individuals being somewhat imperfect rational

economic agents in their retirement and savings decisions.

An illustration of this “desire versus action” compares workers’ attitudes expressed after having

attended an employee education seminar, with actual behavioral changes recorded on company

administrative data systems (Choi et al., 2001a). Immediately following a seminar, for example, all

workers not participating in a firm’s 401(k) plan indicated in a survey that they would join the plan. In

fact, however, over the next six months, only 14 percent did so. A similar, though smaller, gap between

desire and action was true for other behaviors, including intentions to boost saving, change existing

portfolio allocations, or change the mix of future contributions.

Table 1 here

Other Influences. These behavioral findings are further supported by new research on the impact of

investment choices and peer groups on saving rates. One tenet of contemporary economics is that more

choice is better. Yet as Sethi-Iyengar et al. (this volume) show, offering workers many investment

choices can produce “choice overload.” In this case, plan participants become overwhelmed with the

complexity of the decision, and as a result, pension plan participation is reduced. Faced with complex

investment choices, some participants may elect to simplify the decision by following the “default”

heuristic (i.e., “don’t decide, don’t join the plan.”) Similarly, Duflo and Saez (this volume) find that

saving decisions can be strongly influenced by peers. For instance, in several striking experiments and

case studies, they conclude that people with virtually identical demographic characteristics can have

dramatically different saving rates, depending on whether their peers save for retirement or not. They

11

also demonstrate that communications directed to an individual can influence not only the individual’s

savings behavior, but also the behavior of others in his or her work group.

Automatic Saving Plans: Save More Tomorrow. Such behavioral insights into saving behavior have

been illustrated in an interesting way in the Save More Tomorrow (or SmarT) program developed by

Thaler and Benartzi (forthcoming). Under this program, plan participants indicate that they wish to

increase their pension saving rates on a regularly scheduled basis, at prespecified future dates (e.g., on

their anniversary date with the company). This mechanism is designed to address several behavioral

anomalies. First, it recognizes that individuals have self-control problems and benefit from a

precommitment device when it comes to retirement saving. Second, it exploits inertia, since people tend

to sign up initially and the program is automatically carried out in the future. Third, it recognizes the

possibility of hyperbolic discounting: that is, people tend to be averse to saving today but they are willing

to push off their commitment to the future—to promise to “save more tomorrow.” As hyperbolic

discounters, they significantly underestimate the impact of such future commitment. Last, the program

exploits money illusion. Thus participants often think only in terms of nominal take-home pay, so if the

savings increase is designed to coincide with pay raises, they tend to believe that the savings increase had

little or no cost, even though their real current consumption may have declined by a small amount.

In the initial study, the SMarT program was offered to employees at a 300-person firm.

Employees were given the option of financial counseling; most signed up for the counseling, and received

the advice that they should boost their savings rates by an average of 5 percent. Nearly 80 workers took

that advice; many more, just over 160, signed up for the SMarT plan instead, which required annual

increases of 3 percent. After three years, the individuals who signed up for SMarT experienced a

dramatic increase in their savings rates—from 3.5 percent before the plan began, to 11.6 percent (Figure

3).

Figure 3 here.

The popularity of the SMarT program provides further evidence of the divergence between real-

world employees and the rational agents assumed by many economic theorists working on theoretical

12

saving models. Many people attempt to save for retirement and even appear to know when they are not

doing well as they should, but they struggle with exercising the right degree of self-control or willpower.

Through inertia and procrastination, default decisions are easiest to maintain, including saving nothing at

all, or at one’s current rate, particularly if changing behavior requires incurring the costs of saving at a

higher level. Reframing the saving decision to include defaults with automatically higher saving rates,

and using commitment devices, inertia, and money illusion to address the self-control problems of

hyperbolic discounters, all seem useful approaches to address the practical problems associated with the

saving decision.

The Investment Decision

As in the case of the saving problem, the question of how to invest one’s money during the

accumulation phase has been widely explored in a well-developed model of investment decision-making

commonly denominated modern portfolio theory (MPT). The principles of MPT are at the heart of

investment decision-making, both in employer-directed defined benefit plans, and employee-directed

defined contribution plans. MPT influences everything from strategic asset allocation decisions in

defined benefit DB plans, and investment advice and education programs in defined contribution DC

plans, to more technical issues such as performance attribution for investment managers.

In broad-brush terms, MPT seeks to characterize capital market assets, whether stocks or fixed

income investments, in terms of their expected mean return and their volatility or variance—hence the

term “mean-variance” investing. Rational investors seek out efficient combinations of securities that

optimize risk and return, and a given portfolio is on the “efficient frontier” if it offers the highest return

for a given level of risk. Individuals and institutions select from the array of portfolio choices on the

efficient frontier based on their expected utility. In their utility preferences, individuals are presumed to

be risk-averse, meaning that they penalize, or demand higher compensation for, riskier investments. Also,

as risk increases, the compensation they require increases at a faster rate.

13

One of the important predictions of MPT is that investors will be inadequately compensated for

assuming the risks of investing in an individual security. In other words, an efficient capital market will

compensate investors only for the aggregate market risk they endure, so there will be no single-stock

investments on the efficient frontier. Consequently, rational investors will seek to maximize portfolio

diversification and eliminate all stock-specific risk, in the pursuit of optimal portfolio solutions. This

principle has been at the foundation of the growth of low-cost index strategies as an investment

management style in both DB and DC retirement plans.

Another implication of MPT is the theory of time diversification—the closer one is to an

anticipated investment goal where spending from the portfolio begins (such as retirement), the less risky

will be the investment portfolio. In practice, financial counselors frequently propose time diversification

as a popular investment principle, suggesting, in one popular formulation, that people invest 100 percent

minus their age in stocks. The time diversification view is also the basis for most defined contribution

education and advisory services, which suggest that older investors should hold more conservative

portfolios than younger investors. Yet this theory has important critics including Paul Samuelson (1989)

and Bodie (1995), who suggest that investors ought to hold fixed asset allocation percentages over their

entire lifetimes. Finally, richer versions of MPT extend the analysis beyond tradeable securities, to

encompass the people’s broader wealth portfolio. For example, Campbell and Viceira (2002) and Davis

and Willen (2002) suggest that risk and return tradeoffs should encompass illiquid holdings like housing

and human capital.

As with saving theory, behavioral economics asks a very fundamental question about investors in

general, and plan participants in particular: how good are they at actually understanding and acting on the

predictions of mean-variance theory? Arguably, a rational investor should do a reasonable job of

constructing mean-variance efficient portfolios, so there should be some evidence of widespread

competency at these types of investment decisions. Worrisome for the MPT theorists are some key facts

about investor behavior. Of US households who own stocks, the median family owns only two positions,

and even the most affluent households hold a median of 15 (Polkovnichenko, 2003). These low levels of

14

diversification fall well short of the number of positions thought to represent a well-diversified portfolio.

It appears that for many investors, diversification is more akin to holding a variety of assets rather than

the construction of a well-diversified portfolio in an MPT sense.7 A related diversification puzzle is why,

in DC retirement plans, do so many participants over-invest in their employer’s stock? A recent study by

Mitchell and Utkus (2003) estimated that more than 11 million participants held over 20 percent of their

401(k) account in their employer’s stock; of that group, 5 million participants had 60 percent or more in

company stock. Finally, broad stock market fluctuations—like the technology bubble of the late 1990s

and the subsequent bear market—seem hard to reconcile with a model of the investor as a rational, mean-

variance optimizing agent. So do levels of individual and institutional trading in the stock market.

In this section, we first summarize the accumulated evidence on mean-variance behavior among

investors—or rather, the case against mean-variance behavior among investors. Much of this research,

importantly, has been drawn from participants in defined contribution retirement plans in the US. We

then turn to the attempts to develop alternative theories explaining investor behavior.

Lack of Firm Preferences. The findings cited earlier on automatic enrollment illustrate that many

workers lack firm preferences for saving. Merely by rephrasing the question from a positive to a negative

election, workers who weren’t planning to save suddenly find themselves saving—and workers who

would have saved at higher savings rates find themselves saving at the default set by their employer. A

similar lack of strong preferences appears to affect investment decisions. Arguably, if investors were

rational in a mean-variance sense, one would first expect them to have well-defined preferences over their

portfolios. That is, they should have the courage of their convictions. After all, the portfolio they select

represents their unique expectations of risk and return, and it is tailored to their own utility preferences.

In fact, retirement plan participants appear to have relatively weak preferences for the portfolio

they in fact elect (Benartzi and Thaler, 2002). This was found in experiments where workers were given a

choice between holding their own portfolio, the portfolio of a median participant in their plan, and the

portfolio of the average participant: about eight out of ten participants preferred the median to their own.

Only 21 percent continued to prefer the portfolio they initially selected. Furthermore, many found the

15

average portfolio to be quite satisfactory. In other words, pension participants seemed to be quite happy

(or perhaps even happier!) with portfolios constructed at the statistical average of their co-workers’

behavior, than with the portfolios they themselves constructed.

This finding is supported by psychological literature regarding preference reversals. That is,

individuals often do not arrive at a decision with firm preferences in mind; preferences appear not be

hard-wired. Rather, individual preferences tend to be situational and emerge at the time a decision in

made, based on the conditions and information surrounding that decision. To the extent this is true,

preference reversals tend to be more common than might be expected. Individuals who thought not to

save find themselves saving; individuals who selected their own portfolio find themselves just as happy,

if not happier, with another choice.

Framing Effects. Just as saving choices can be affected by framing, so too can investment decisions be

influenced, sometimes strongly, by framing effects. Much of the research in this area has investigated the

impact of investment menu design on participant investment choices in defined contribution retirement

plans. The theme underlying this research is that menu design is a more powerful influence on participant

decision-making than the underlying risk and return characteristics of the investments being offered. In

this sense, the investment menu in a retirement plan is an “opaque” frame, which most participants cannot

see through, to understand the underlying risk and return characteristics of their investments. Put another

way, many participants appear to have weak convictions regarding risk and return, and they can easily be

swayed in their decisions by the framing effects of an investment menu.

In one experiment, participants were asked to select an investment mix for their retirement plans

given two fund offerings (Benartzi and Thaler 2001). Some participants were presented with a stock fund

and a bond fund; others with a stock fund and a balanced fund; and a third group with a bond fund and a

balanced fund. In all three cases, a common strategy was to choose a 50/50 mix of the two funds offered,

although many participants did select different weightings. What was striking in the data was fact that

radically different underlying asset allocations ensued, given the different choices offered. For people

given the choice of an equity fund and a bond fund, the average allocation to equities was 54 percent. For

16

those offered two equity-oriented portfolios, a balanced fund and an equity fund, the average allocation to

equities was 73 percent. And for those offered a balanced and a bond fund, the average allocation to

equities was only 35 percent. In a related experiment using investment menus with five funds, the

authors found that the asset allocations chosen by participants were again strongly influenced by menu

design. If the plan offered several equity funds, participants invested more in equities; when it included

more fixed income funds, they chose fixed income options instead.

A different study also asked plan participants to select investments from three different menus,

which the authors posed might be similar to the structure of a privatized Social Security account (Benartzi

and Thaler 2002). The investments allowed ranged from A (low risk) to D (high risk). The first menu

offered included options A, B and C; the second menu, just options B and C; and the third menu, options

B, C and D. Comparing options B and C, which were in all three menus, 29 percent of the participants

preferred C over B in the first menu; 39 percent in the second menu; and 54 percent in the third menu. In

other words, in the first menu, where option C was at the extreme, it was liked least; in the third menu,

where option C was the middle choice, it was liked most. As with the asset allocation experiment above,

this shows that participants appeared to use a naïve heuristic (i.e. “avoid extremes, pick the middle

option”) rather than maintain a consistent set of well-ordered risk preferences to select from the

investments offered.

Related research indicates that, beyond these menu effects, even simple changes in the way

information is presented can influence asset allocation decisions (Thaler and Benartzi 1999). In one

experiment, plan participants were asked to make investment decisions based on reviewing the one-year

return profile of US common stocks; in a second experiment, they made decisions based on a 30-year

return profile. In the first instance, the average allocation to equities was 63 percent; in the second, 81

percent. The implication is that plan sponsors can alter asset allocations if return data are presented over

different holding periods. And as Scott and Stein show (this volume), different types of investment

education and information can substantially change retiree investment allocations.

17

Like the saving research discussed earlier, these findings underscore the powerful influence of

framing effects on decision-making in retirement plans. Apparently, many plan participants seem to lack

well-formed investment preferences, and these preferences appear to be easily altered by the way the

choices are presented to them.

Inertia and Procrastination. As with savings behavior, inertia also plays a large role in investment

decision-making, in addition to these framing effects. Madrian and Shea (2001) and Choi et al. (2001b)

reported high levels of inertia in investment decision-making in their studies of participants and automatic

enrollment. To further underscore this point, we examined how 2.3 million plan participants at the

Vanguard Group allocated their new contributions accounts as of June 30, 2003 (Figure 4). First, we

found that fewer than 10 percent of plan participants change their contribution allocations each year.

Further, participants who initially enrolled in their plans near the top of the bull market in 1999, allocated

about 70 percent of new contributions to equities in June of 2003, notwithstanding the huge market drop

sustained over the preceding three-year period. Meanwhile, participants who newly enrolled during the

first six months of 2003, after the three-year fall in US equity prices, allocated only 48 percent of new

monies to equities. While this illustrates how sensitive participant investment decisions at enrollment are

to then-current market conditions, it also demonstrates the power of inertia. It seems unlikely that

participants enrolled in 1999 would have dramatically different risk preferences than those who enrolled

in 2003, yet the recent enrollees were presumably making active choices based on then-current

information, whereas earlier enrollees did not react so dramatically to market news.

Figure 4 here

Figure 4 also illustrates “anchoring” effects for pension investors. Anchoring refers to the notion

that decision-making is strongly influenced by starting values, no matter how arbitrary they may be.

Among participants, it appears that the relevant anchor is their initial allocation decision, and subsequent

portfolio changes tend to be made with reference to that initial value, rather than on some absolute basis.

For instance, participants who enrolled at the peak of the bull market continued to allocate seven out of

18

ten dollars to equities by 2003, over twenty points higher than those enrolling in the first six months of

2003.

The Conundrum of Employer Stock. The use of company stock within US defined contribution plans

offers a compelling case study about the relevance of mean-variance models to investor decision-making.

As noted earlier, Mitchell and Utkus (2003) have calculated that 11 million plan participants have

allocations above 20 percent of their account balance in company stock—and 5 million have allocations

above 60 percent of their account balance. A conventional economic explanation for this phenomenon is

that employers and stockholders seek to promote employee productivity through stock ownership, and so

they encourage or mandate large employee holdings of company stock. As rational agents, however,

employees who are aware of the risks they are being required to assume, should demand compensation in

some other form, such as higher wages or benefits.

There is some support for the “rational agent” view of workers holding company stock. This is

because concentrated company stock positions are most common for large firms, and such firms typically

pay higher wages and benefits to their employees. Yet from a behavioral perspective, there is also

evidence that concentration stock positions are not solely due to incentive effects; rather, it seems that

computational or behavioral errors on the part of participants also help explain the phenomenon. For

instance, Mitchell and Utkus (2003) use survey data to uncover evidence of “risk myopia” regarding

employer stock, in that many participants rate their employer’s stock as safer than a diversified equity

fund.

Another Vanguard survey (Table 2) illustrates that even after the post-Enron publicity

surrounding company stock, two-thirds of participants rate their employer stock as safer than, or as safe

as, a diversified portfolio of many stocks. Only one-third said it was more risky. What is striking about

these results is the comparison between participant risk perceptions and the actual return and volatility of

their employer’s stock. Looking at the risk ratings first, it is natural to conclude that at least two-thirds of

participants are not mean-variance investors when it comes to company stock. They rate stock as safer

than or as safe as a diversified portfolio, despite its actual higher volatility than a broad market index: a

19

clear-cut “error” under modern portfolio theory. Arguably one-third of participants did assess the risk

correctly, in that they rated their employer’s stock as riskier and its volatility higher. But it seems

implausible to conclude that all participants who understand mean-variance analysis may only be found

among the set holding riskier employer stocks; it is more likely that participant do not base their risk

perceptions on volatility. Instead, participants’ risk ratings are well-correlated with the historic relative

returns of their employer’s stock.

Table 2 here

The conclusion that plan participants overlook volatility and focus on returns is supported in

Benartzi’s (2001) study of pension investments in employer stock. Specifically, he finds that participant

allocations were based on extrapolations of the company’s historic stock performance. Participants who

overweighted their employer’s stock based on good past performance then found that those stocks

subsequently generated below-average performance. Conversely, those participants who underweighted

their employer’s stock due to poor past performance subsequently saw the stock becoming an above-

average performer.

Participants’ allocations were also influenced by whether their employer provided a match in

company stock, a phenomenon that Benartzi dubbed the “endorsement effect.” The conclusion is that,

just as menu design influences participant investment decisions, so too does the employer’s plan design

decision. Offering a match in company stock encourages participants to hold more in stock than workers

whose employers do not match in stock. Other researchers have also argued that past performance, rather

than risk, drives participants’ portfolio decisions (e.g. Choi et al., this volume; Huberman and

Sengmueller 2003; Poterba et al., 2003; Purcell, 2002).

Reliance on Past Performance. Why do investors irrationally rely on past performance and fail to take

expected returns as well as risk into account, as modern portfolio theory suggests they should? Two

behavioral phenomena may offer some answers. A first issue is the pervasiveness of the

“representativeness heuristic” in decision-making, explored by Tversky and Kahneman (1974). They

found that people tend to see patterns in small series of randomly drawn numbers, and when making

20

decisions, people attempt to impose some order or structure on the information that they see. For

example, mutual fund investors might identify a fund manager with three years of top performance and

conclude that the manager has unusual skill—rather than view it as a random process. Of course, viewed

across the universe of thousands of investment managers, a given manager’s three-year track record is just

as likely an indication of chance as of skill. The representativeness bias may partly arise due to a framing

problem: that is, rather than frame the skill versus luck decision in terms of the universe of all individuals

making portfolio decisions, a fund investor may frame it narrowly in terms of the three-year track record

of a single investment manager. As a result, what may actually be a random outcome may instead appear

to be logical sequence.

A second issue is that many people appear to be subject to what has been called an “availability

heuristic”: when faced with difficult decisions, they tend to rely on readily available information. A

simple reason that investors may rely on past performance could be because that information is cheaply

available. As any plan participant knows, retirement plans and investment companies generate prodigious

amounts of past performance data which they make available in statements, on websites, in enrollment

materials, and in newsletter updates. Past performance is also pervasive in the media, but very few report

systematically on expected returns. Of course in the US and elsewhere, reports on past investment

performance are often accompanied by the legal disclaimer that “past performance is no guarantee of

future results.” Yet one need only compare the size of that disclaimer to the volume of past performance

data to understand its limitations in the face of the availability heuristic.

In sum, the representativeness and availability heuristics may help explain why, for example,

mutual fund investors invariably chase performance in their fund purchase decisions (Patel et al., 1991).

Patterns suggesting superior performance are constructed from small samples drawn either from skill or

luck. And the pervasiveness of past performance data leads to an inevitable reliance on past performance,

despite the legal caveats.

Prospect Theory. If plan participants are not necessarily mean-variance investors, then how do they

actually make decisions under uncertainty about their pension investments? This is a complex question,

21

partly because behavioral research in the last 20 years has focused on analyzing how people evaluate

risky choices in general; only recently has interest has turned to investment decisions.

The seminal theory of risky decisions was offered by Kahneman and Tversky (1979), who

modeled individuals as though they made decisions to maximize an S-shaped value function depicted in

Figure 5. This function differs from conventional utility maximization in two critical ways. First,

individuals are thought to consider not how a decision influences total wealth (as in standard utility

theory) but rather incremental gains and losses. Second, individuals are modeled as treating gains and

losses quite differently. The gain function (to the right of the origin) is concave, while the loss function

(to the left of the origin) is convex, with a much steeper slope. What this means is that individuals will

experience losses more acutely than gains for a given dollar of gain or loss. Their experimental evidence

suggested that the index of loss-aversion is about 2.5: in other words, when evaluating risky gambles, the

individual will report that losses are 2.5 times as painful as the equivalent dollar value of gains. For

example, if someone were presented with a 50 percent chance of losing $1,000 or a 50 percent chance of

gaining an unknown amount, $X, the evidence suggests that many people would not entertain this gamble

until the value X is on the order of $2,500 or so.

Figure 5 here

This approach has come to be known as prospect theory, and it has potentially important

implications for investment behavior. For instance, investors will seek to lock in certain gains and avoid

certain losses. This implies that individuals tend to be risk-averse for a known gain, but they can become

risk-seeking in an effort to avoid a certain loss. In addition, actual behavior will depend on the exact

sequence of gains and losses and how the individual has incorporated prior gains and losses into current

perceptions. For example, suppose an individual wins $100. If offered a reasonable chance to win more

money or lose the $100, many people would decline the additional gamble, because of the risk of

forfeiting the $100 sure gain. But if offered a choice to win more money while preserving a meaningful

part of the $100 gain, many people take the risk. This is known as the “house money” effect: while

people are generally risk-averse in the domain of gains, if they feel they are risking someone else’s money

22

(e.g. accumulated earnings from prior bets), they become more risk-seeking. On the loss side of the

equation, after losing $100, many people will accept a gamble that entails losing significantly more than

$100 in an effort to recoup the $100 loss. This represents both the element of risk-seeking in the domain

of losses, and the “breakeven” effect. Faced with the realization of a certain loss, many people seek

additional risk, in an effort to recoup their investment, contrary to the conventional economic notion that

“sunk costs are sunk costs” (i.e. rational agents should ignore realized losses). This approach offers an

explanation for why investors have difficulty realizing losses on their investments (a strong desire to

avoid loss realization and break even). It also might help explain why they sometimes increase risk-

taking in risk equity markets (existing gains appear to be locked in and are “house money” which can be

gambled”) and in falling markets (existing losses appear temporary and extra risk-taking will help recoup

those losses).8

Prospect theory and behavioral economics have been deemed relevant to investment decision-

making in three ways (Kahneman 2003). First, in the area of gains, investors are often characterized by

overconfidence and excessive optimism: people tend to construct forecasts of the future that are typically

too rosy. Second, in the area of losses, investor risk-aversion will lead to an unwillingness to realize

investment losses, and a premature realization of investment gains (called by Shefrin and Statman the

“disposition effect”;1985). And third, if decisions are less than optimal due to both overconfidence and

loss avoidance, the impact of these anomalies will be exacerbated by narrow framing effects. We

consider each of these elements in turn.

Overconfidence. In the domain of gains, one of the important findings of psychology and behavioral

economics is that peoples’ future forecasts are often characterized by widespread overconfidence and

excessive optimism. Overconfidence is a widely documented trait in human decision-making: thus

people systematically overrate their skills on such parameters as driving skills and humor. In business,

managers tend to be overconfident about their abilities: for example, managers tend to overpay for

mergers and acquisitions because of overoptimistic assessments. Other professionals in a range of diverse

fields—in psychology, medicine, investments, engineering, and so on—have demonstrated

23

overconfidence in their decision-making. In a study on future life prospects, college students were asked

to evaluate the chances of certain positive and negative events occurring in their lives (e.g., having a

bright career versus experiencing professional failure; maintaining ongoing good health versus

contracting a mortal disease while young; having a happy versus a difficult domestic and emotional life).

When asked to judge their own prospects, the students downplayed negative life events and emphasized

positive outcomes. What was interesting was that, when asked the same questions of their college

roommates, the students were more even-handed in their responses. Similarly, in medical decision-

making, patients with mortal diseases were shown to be much more optimistic about their future

prospects than their professional caregivers. Overall, the accumulated psychological evidence regarding

overconfidence in decision-making has been described as the “Lake Wobegone” effect, named after a

fictional US town described on a popular radio program where “all of the children are above average.” 9

Such overconfidence may partly be the result of an inability to understand accurately the role of

random chance in determining the future. As noted above in our discussion on representativeness, people

are notoriously poor statisticians, and they find patterns and trends in data that could just as easily be

explained by random chance. Individuals appear to significantly underestimate the impact of random

chance on their lives, and in hindsight overemphasize the degree of control they have over outcomes.

Lack of objectivity might help explain self-evaluations: Individuals generally perceive themselves as

better than others—and have better views of themselves than others do. Poor risk calculations certainly

play a role in overconfidence: individuals who are “100% sure” of their responses to certain questions are

usually wrong 20% of the time. Perceived sense of control also plays a role: the stronger one’s sense of

control, the more powerful one’s sense of confidence. Asked about the risks of a disabling car accident,

people are much more optimistic when they are driving, than when they are passengers. There is also a

gender element at work: men tend to be more overconfident than women.10

Overconfidence probably has some economic, psychological, and even evolutionary positive

benefits: for example, it may be the well-spring of risk-taking and entrepreneurial activity, or it may allow

fast recovery from life’s disappointments. But in the domain of investing, it may also lead to behaviors

24

that are less than optimal and certainly at odds with mean-variance theory. For instance, overconfidence

helps explain the high levels of trading activity in equity markets. Barber and Odean (2000) report high

rates of turnover, on the order of 75 percent, among households owning brokerage accounts. They

calculate that trading is typically hazardous to one’s wealth, with active traders earning 11.4 percent over

a five-year period, while the market returned 17.9 percent, and low-turnover accounts 18.5 percent. That

study also reports that men trade 45 percent more than women and the difference is even stronger for

single men versus single women. Agnew et al. (2003) suggest that these trading results may be less

relevant for retirement plans, since participants could mentally account for brokerage and retirement

investments differently.

Other research focuses on affluent male equity investors, and it has confirmed that

overconfidence and an emphasis on personal skill are both important attribute of investors, at least in the

sample of experienced investors surveyed by De Bondt (1998). Most of these investors exhibited a high

level of confidence about the stocks they had selected, while their outlook for the broad market was not as

positively biased. They relied on their own skills at selecting companies and tended to downplay the

importance of modern portfolio concepts like diversification. From their perspective, it seemed that

diversification was less focused on non-correlated stock holdings and more about simple variety. These

investors also tended to be surprised about the relationship between the overall market and their own

holdings: that is, they downplayed the impact of market forces on their portfolios.11 These diversification

results seem consistent with Goetzmann and Kumar’s findings (2001) that individual investors do own

multiple stocks, but these are not drawn from uncorrelated industries and sectors so they are not typically

diversified in the modern portfolio sense. Statman (this volume) shows how behavioral theories of risk-

taking may be used to formulate alternative theories for portfolio construction.

Loss Aversion and the Disposition Effect. If overconfidence helps explain behavior on the “upside”

side of the prospect theory ledger, then the “downside” is dominated by aversion to loss realization. This

plays out in interesting ways. For instance, as noted above, people are inclined to take a gamble if

confronted with the choice of realizing an incurred but not-yet-realized loss, versus taking the gamble in

25

which they might break or lose more. Particularly if there is a reasonable prospect of breaking even and

avoiding a loss, many people take the gamble and risk losing even more money. In the investment

setting, this manifests itself in what Shefrin and Statman (1985) call the “disposition effect.” People who

invest in stocks appear to rush to realize gains too quickly: they try to lock in or make certain the gains

that they have already realized. On the other hand, they also appear to have trouble “cutting their losses;”

that is, they hold onto loss-making stocks too long in the hope of recovering their investment. The impact

of this effect on brokerage account investors is not small: Odean (1998a) calculated that investors who

sold winning stocks saw those stocks outperform the market by 2 percent in the subsequent year, while

investors who kept their losing stocks saw those stocks underperform by market by 1 percent over the

same period. In total, the net impact of selling winners too quickly while holding losers cost investors 3

percent per year in terms of portfolio returns. These results also validated the notion of overconfidence in

investment decisions: people continued to hold and to sell the wrong stocks, leading to lower returns,

despite actual results.

Narrow Framing. Adding to what has already been discussed, there is evidence that overconfidence and

loss-aversion are exacerbated by too narrow a framing of risky decisions (Kahneman 2003). Few

investors would take a gamble involving a 50 percent chance of wining $1,500, versus a 50 percent

chance of losing $1,000. (With loss-aversion parameters of around 2.5, most investors would not take the

gamble until the gain was closer to $2,500.) Yet experimental evidence indicates that people are more

willing to accept this gamble when they are given the opportunity to play it many times, or when it is

framed in terms of changes to their entire net worth. Perhaps, as Kahneman observes, it is more natural

for investors to “think small” when facing a one-time gamble, but they may get it right and “think large”

if facing sequences of gambles or changes to total wealth.

The Decumulation Decision

The last phase of financial decisionmaking for retirement happens during the decumulation

period. This is likely to occur during later middle-age and beyond, and it is the period when most people

26

decide how they will spend down their accumulated assets. Of course, if there were no uncertainty, the

rational life-cycler would plan to spend down retirement assets so as to ensure optimal retirement

consumption and protect bequest motives (if any). People having saved for retirement via a formal

pension scheme would be expected to decumulate their assets just as their non-pensioned counterparts

having the same total assets, except insofar as it reduced the retiree’s tax obligation (Brown et al., 1999).

In practice, of course, people confront many sources of risk during the retirement period. The

most important of these are longevity risk, inflation risk, health risks (leading to unexpected expenses and

costs), and capital market risks. All or a combination of these risks can contribute to experiencing

consumption shortfalls during retirement—or simply running out of money. So many fundamental

uncertainties, further complicated by the psychological considerations discussed above, combine to make

it quite difficult for retirees to deftly manage the drawdown process for retirement accounts in old age. In

this section, accordingly, we first summarize available evidence on how people deal with longevity risk,

then turn to a discussion of inflation risk, and finally conclude with a brief discussion of how to manage

capital market risk during the withdrawal period.

Longevity Risk. Because people do not know precisely how long they will live, they run the risk of

exhausting their assets before dying. Such risk exposure can be reduced by consuming less per year

during retirement, but of course this simply elevates the chances that a retiree might die with “too much”

wealth left over.

One way to offset longevity risk is to buy an annuity with all or part of one’s retirement assets

(Brown et al., 2001; Mitchell and McCarthy, 2003; Sondergeld and Drinkwater, this volume). Single

premium lifelong annuities are relatively appealing, since they continue to pay benefits as long as the

retiree lives, irrespective of whether the retiree outlives the life tables. Indeed, recent survey analysis

(Panis, this volume) finds that retirees holding annuities are more satisfied with their retirement, holding

other things constant. Consequently, the implication is that at least partial annuitization may provide

peace of mind associated with longevity protection.

27

Notwithstanding the substantial theoretical appeal of annuities, however, relatively little

retirement money is currently devoted to the purchase of annuities in most developed countries. For

instance, life annuity purchases in the US amounted to more than $120 billion in 1999, but the majority of

sales were for variable annuities which are used mainly in the accumulation process rather than for

decumulation products paying lifetime benefits (Brown et al., 2001). In the group pensions market, there

is also growing attention to this issue. Previously, defined benefit plans normally paid either single or

joint and survivor life annuities as a matter of course; rarely was any sort of lump sum option available in

lieu of the lifetime benefit stream. Of late, however, defined benefit pensions have begun offering lump

sum distributions to their retirees, akin to the payouts popularized in the defined contribution world

(Moore and Muller 2002). As a result, workers reaching retirement age with pension coverage are

increasingly unlikely to take their benefits as lifetime annuities. Indeed, a recent study found that three-

quarters of company pension distributions are currently paid as lump sum cashouts rather than as lifetime

annuity payments (McGill et al. forthcoming). In this sense, fewer and fewer retirement plans are

providing longevity insurance in the form of lifetime insured annuity benefits.

Several explanations for the declining demand for annuitization in retirement have been offered.

One factor is that people may be poorly informed regarding their remaining life expectancies, tending to

underestimate the risk of outliving one’s income. For instance, a recent industry survey reported that only

one-third of the respondents knew that someone who attained the age of 65 had a substantial chance of

living beyond his life expectancy (Metlife 2003). Yet other surveys report that older people’s expected

survival patterns track actuarial tables relatively closely (Hurd and McGarry 1995), and retirement asset

shortfalls appear uncorrelated with people expecting to die soon in retirement (Mitchell et al. 2000). A

different factor discouraging annuity purchase is that retirees often have strong bequest motives, and

many of them expect to have to pay for long-term care. In such cases, they might elect to hold on to their

funds rather than annuitize them on retirement. And of course, to the extent they have Social Security and

defined benefit pension plans, they will be less likely to annuitize all their assets since they are partly

protected against longevity risk already (Brown et al., 2001).

28

Three other “rational” explanations may provide insights into why annuity purchases are low

despite the fact that baby boomers are moving into retirement age. First is an interest rate factor. The

decision to annuitize at a given point in time represents an irreversible decision to “lock in” then-current

yields (which underlie the contract pricing). A second is the cost factor: retirees sometimes see insured

products as uncompetitive with pure investments due to the loads levied by the insurance providers. Yet

the loads have decreased substantially over time, and evidence indicates that retirees can expect high

“money’s worth” for annuity products in many countries (Mitchell et al. 2000). Consequently the

respectable returns combined with the insurance protection should induce more interest in this payout

structure as the baby boomer generation moves into the retirement years. The other main reason lump

sums are attractive is that regulations currently permit workers to take a relatively large lump sum

computed with a transitorily depressed discount rate, and in many cases this is more economically

profitable than leaving the funds in the plan to grow.12

Behavioral factors may also explain the low demand for annuities in retirement, including most

importantly loss aversion. This arises because some retirees may worry about potential losses to heirs in

the event that they die “early”, since annuitization typically eliminates the possibility for bequeathing

these funds. Adding to the problem is that retirees may heavily discount future benefit coverage in the

event that they life a long time in retirement. Such an asymnmetric valuation could enhance peoples’

probability of taking their pension accruals as a lump sum versus buying a life annuity, and may explain

why some argue that “locking up” one’s assets in annuities boosts rather than reduces risk. To meet such

concerns, some insurers have begun to combine annuity offerings with life insurance, long-term care, and

disability benefits, so as to reduce the fear of “losing it all” due to premature death. Employers wishing to

help workers with their self-control problem might offer annuities as the default option at retirement,

rather than making the lump sum the standard choice.

Nonetheless, it does appear that many people fundamentally undervalue the appeal of a lifetime

annuity—sometimes at substantial, if not overwhelming, cost. One fascinating example is a study of

annuitization behavior for personnel at the US Department of Defense (Warner and Pletter, 2001). In

29

1992, about 65,000 officers and enlisted personnel were involved in a program to reduce staffing at the

Department of Defense. To this end; they were offered payments from their retirement plan in the form

of an annuity or a lump sum. The internal rate of return on the annuity ranged from 17.5 percent to 19.8

percent, at a time when government bond rates were around 7 percent. Economists estimated that all of

the officers and half of the enlisted personnel would take the annuity. In the end, contrary to

expectations, 52 percent of the officers and 92 percent of the enlisted personnel took the lump sum. In

total, the DOD employees forfeited a total of $1.7 billion in economic value, by electing the lump sum

over the annuity.

Inflation and Capital Market Risk. Last, but certainly not least, we turn to the risk of inflation and

capital market risk during the retirement period. It is somewhat well known that the common worker is

rather poorly informed about volatility in asset returns and inflation rates (Bodie et a., 2002), a problem

that also besets him during the retirement period as well. For instance, from the late 1970’s to the late

1990’s, the US had a relatively low rate of inflation and rising stock prices that contributed to a

widespread belief that equities serve as a good hedge for inflation. Yet this is not accurate: during the

1970s, inflation moved into double digits, yet stock prices fell by over half in a short two-year jolt (1974-

5). Brown et al. (2000) also confirm that stocks have been a poor inflation hedge in the US, at least in the

short- to medium-term. For this reason, retirees seeking protection against the destructive impact of

inflation over a long retirement period would benefit substantially from holding at least a part of their

financial assets in inflation-protected assets such as Treasury Inflation-protected Securities (TIPS).

The fact that workers take lump sums from their pensions, rather than have their funds continue

to be managed by the pension fund itself, may also be of concern for several other reasons. One potential

explanation is overconfidence: many people believe they can live well on relatively small asset pools

during retirement, yet after leaving work, they then find they run out of money, sometimes within a few

months of retirement. This is exacerbated by the lump sum benefit is often “framed” in a way that induces

them to overvalue the lump sum and undervalue the annuity. Offering a retiree a lump sum of $100,000,

30

versus taking a joint and survivor annuity of $600 per month for life, tends to highlight the “massive sum”

versus the longevity protection.

Another concern is that, due to lack of self-control, people may be tempted to spend their lump

sums once they are out of the pension plan. While recipients do roll over some of their lump sum

distributions into Individual Retirement Accounts, they do use a large fraction to pay debts and cover

current expenses (Moore and Muller 2002). A problem with taking a lump sum from the pension, of

course, is that the amount withdrawn becomes subject to regular income tax and it may also elicit excise

tax if the recipient is younger than age 59 ½. Finally, many retirees are poorly equipped to manage their

investments in old age, perhaps because they never were particularly financially literate, or perhaps they

suffer diminished faculties due to poor health and lack of mobility with age. And it is difficult even for

experts to undertake the sophisticated calculations required to simultaneously manage the investment

portfolio, the drawdown rule, and the target horizon over which these decisions are made (Dus, Maurer

and Mitchell 2003). Most financial planners are not particularly well versed in these techniques either, nor

are their clients. But possibly due to overconfidence, they often expect the money to last longer and earn

more than a prudent strategy would dictate.

Policy and Plan Design Alternatives

Our overview thus far has illustrated how behavioral economics and finance research of the last

few years has fundamentally challenged the ways in which plan sponsors, retirement service providers,

and policymakers should think about retirement plan design in the future. At the most inclusive level,

behavioral research offers several new insights about the nature of individual decision-making in

retirement plans, which we enumerate next. After noting these, we turn to several plan design and policy

responses to this new body of research.

1. One lesson is that behavioral research challenges some of the most central assumptions of decision-

making. In particular, it challenges the notion that workers are rational, autonomous, microcalculators

who exercise independent and unbiased judgment when it comes to their workplace retirement plans. The

31

evidence suggests that people do strive to maximize their self-interest, but for a variety of reasons

outlined here, they often fail to act in accordance with the expectations of rational economic and financial

theory, in both the accumulation and decumulation phases. Some people have self-control problems

when it comes to saving; such individuals could benefit from commitment devices. Others simply

overdiscount the future and overvalue the present; such people could benefit from precommitment

pension savings programs. Still others might be unduly influenced by defaults and inertia: as a result,

their attitudes and action diverge so they very much want to save more for retirement, but they do not.

Here too, precommitment devices can plan an important role. Also, some individuals do not appear to

evaluate their investment portfolios in mean-variance terms; that is, past performance and risk errors

cloud their judgment. They are overconfident about the future and have trouble cutting their losses.

Some trade too much, think “too small” in terms of gains and losses, and take lump sums when in fact all

of these behaviors increase rather than reduce risk.

2. Another lesson learned is that plan design drives participant decisions, often in unanticipated ways.

Behavioral finance and economics also challenge the notion that pension plan design is a neutral vehicle

within which participants make their own choices independently. Because of default, framing, and inertia

effects, we have argued that the design of a retirement system or plan has a profound effect on participant

investment and saving decisions. Sponsors and policymakers can alter behavior in fundamental ways by

choosing different default structures. In particular, the design decisions to set up automatic enrollment,

automatic saving, or default investment programs, which makes some saving and investment decisions

automatic, are particularly critical.

Using traditional policy and plan design language, defined contribution plans are “employee

directed,” with employees seen as the active agents while the employer is thought to play a minimal

decision-making role. In some sense this is a libertarian decision-making model, where independent

agents can act to maximize their personal welfare within the constraints of the system. But behavioral

research sketches a different picture of many workers. These are people with weak or uncertain

preferences about basic questions as how much to save, or how much risk to take. Plan design decisions

32

then emit powerful signals about “appropriate” employee behavior, and employer/policymaker design

specifications trump independent decision-making. Such an environment is consistent with the

“paternalistic libertarianism” notion advanced by Thaler and Sunstein (2003), where individuals can be

offered choice, but paternalistic elements of retirement plan design play a powerful role in shaping the

choices offered.

3. A third finding is that the standard approach taken in most contemporary defined contribution plans

may be counterproductive in encouraging retirement saving. Generally, participants are told that (1)

saving for retirement is optional (since joining the plan is discretionary); (2) the need to increase saving

over time is optional (it requires a voluntary election by the worker); and (3) investing for retirement

should focus on principal stability, rather than taking on risk or balancing the portfolio (since the default

fund in most retirement savings plans is a conservative fixed income option). It is interesting that

employers and policymakers rely on a model of voluntary choice by the worker in retirement saving

plans, although they do not do so in other components of workplace benefits programs. Here a

comparison is instructive between employer retirement saving plans and health insurance. In terms of

health plan participation, it has been uncommon for a US employer offering health insurance to allow

employees to drop coverage altogether (particularly without proof of some other health insurance).

Employers also regularly make decisions about healthcare plan coverage levels and, for example, types of

procedures and catastrophic coverage, rather than leaving such choices to voluntary election by the

worker. Perhaps this is because lack of health care coverage might be felt immediately in the workplace

if a worker became ill, while low levels of retirement saving have no immediate consequence. If this is

true, employers maybe subject to the same type of “hyperbolic discounting” as many workers,

overvaluing present-day risks, and overdiscounting future concerns.

4. A fourth lesson is that the current education model in 401(k) plans may have reached its effective

limits. Accompanying the growth of participant-directed DC retirement plans has been a large expansion

in the provision of workplace education. Much of the educational effort has been motivated by

nondiscrimination testing—employers have an incentive to encourage plan use among lower-paid

33

employees, to allow highly paid employees greater ability to contribute to the plan. Other motivations

have been employers’ desires to promote a popular saving benefit, and to minimize fiduciary liability for

participant investment decisions. The current educational model tends to emphasize communication and

education activities, both of which are aimed at producing behavioral change (e.g., joining the plan,

boosting saving, investing more effectively). Yet the behavioral literature suggests that for many

workers, this model is limited in its applicability. There is the problem of inertia, which we have

described as the divergence between desire and effective action. There is also the notion that only part of

the workforce is motivated to learn about personal finances or interested in using financial education.

Contemporary education practices assume that most workers are rational agents and planners, but the

evidence we offer suggests that large numbers of workers simply are not.

As Selnow (this volume) suggests, an alternative model is that desired behavior must precede

education. Mechanisms must be found, whether through plan defaults or delegation to a third party,

where workers begin practicing the right behaviors at the outset. Education then can play an ancillary

role, explaining the rationale for the defaults and alternative courses of future action. In effect, behavioral

economics suggest a reversal in the causality of education: a shift from education driving behavioral

change, to initial behavioral change preceding education.

These broad themes—imperfect investors and savers, the critical role of system design, a new

model for education—suggest a number of policy and plan design choices. Four issues deserve

prominent attention:

1. Much depends on the default contribution arrangement in defined contribution plans. One way to

exploit the findings of this rich new behavioral literature would be to alter the nature of default decision-

making in US retirement saving plans. Inertia, procrastination, and lack of decision-making willpower

can be exploited to encourage more retirement saving. An “auto-pilot” 401(k) is one possibility. Also

automatic enrollment of all eligible employees; scheduled annual savings increases (as in the Save More

Tomorrow plan); and a selection of default investment choices that represent optimal portfolio choices,

such as a series of age-based balanced portfolios. In this way, the passive decision-maker may rely on

34

system design to reach a near-optimal retirement outcome. While workers will still retain the right to “opt

out” of this arrangement, allowing for freedom of choice, the system design always directs workers

toward desirable saving and investment behaviors.

Auto-pilot 401(k) solutions are not without their drawbacks: most notably, the automatic

enrollment of employees will raise costs above the current model where workers must opt-in (e.g., higher

employer matching contributions, higher administrative costs for employers and providers). Offsetting

these costs, in part, could be the greater asset pool resulting from higher contribution rates over time, as

well as higher-fee investments as default options. Just as employees are easily influenced by employer

plan design decisions, so employers are influenced by implicit and explicit policymaker directions.

Currently, clear-cut regulatory guidance exists only for automatic enrollment components; to generalize

this success, some type of regulatory or statutory endorsement for automatic saving and age-based

investment choices would be required. A further consideration, of course, is whether the auto-pilot 401(k)

model would be an alternative to existing nondiscrimination testing requirements. From a policy

perspective, nondiscrimination testing rules were introduced to ensure that low-paid workers take

sufficient advantage of tax-deferred retirement savings plans, and that tax benefits of such plans did not

accrue solely to the highly-paid. An auto-pilot 401(k), perhaps with some standardized eligibility and

matching contributions (as in today’s “safe harbor” design), might accomplish the same objective in a

different way. It would offer other benefits for lower-paid workers as well—including automatic

retirement savings increases and better portfolio choices for many.

2. Simplified menu design in retirement plans could be very useful. One of the more practical conclusions

from behavioral finance is that investment menu design must be closely scrutinized. Iyengar et al. (this

volume) describe the “choice overload” hypothesis in detail: complex investment menus may discourage

plan participation. But it is also clear that even simple pension design decisions, such as the composition

of equity versus fixed income funds, can also strongly influence participant investment behavior.

Certainly, one implication from the research is that many participants lack skills needed to make complex

investment choices among highly differentiated options. There is little evidence that participants are

35

constructing the mean-variance optimal portfolios that employers use to justify the inclusion of different

investment classes, styles and managers. Instead they use informal heuristics, including past

performance, to make choices.

As a result, the research suggests that the “laundry list” approach to investment options – where

workers are given 50 or 100 choices of funds – will be confusing and demotivating for some.

Accordingly, plan sponsors might experiment with tiered investment choices, where communication

resources are devoted to educating participants about a limited menu of core options, and additional

choices for sophisticated investors could be segregated from the core menu.

More broadly, both employers and policymakers need to rethink the trend toward expanding the

myriad and complex active saving decisions presented to workers. Behavioral research suggests that

there are natural, inevitable limits to a policy of ever-increasing choices and decisions. For example, tax

legislation in 2001 authorized an after-tax version of 401(k) saving plans (dubbed a “Roth” provision, in

which contributions are made after-tax and all future earnings and distributions are tax-free). The idea is

that employers could offer employees a choice between pre- and post-tax savings in the same plan. The

Administration has also recently proposed to create a system of after-tax personal saving accounts which

compete with pre-tax workplace saving plans. Whatever the merits of these proposals, it seems likely that

adding new choices will further complicate investment and saving decisions. Since the evidence suggests

that many workers already struggle with the basic decisions to save, invest, and spend during retirement,

it seems likely that new options will further challenge already-burdened decision-makers.

3. New approaches are needed to help workers and retirees better manage company stock risk. Congress

and employers have attempted to address the risks of holding excessive company stock through education

and educational/disclosure activities, yet the behavioral evidence suggests that this strategy will have

limited impact. One problem is inertia; reducing concentrated stock positions requires taking a

disciplined approach to selling stock holdings, but few participants tend to follow such a self-motivated,

disciplined approach to managing their saving due to inertia and procrastination. Another problem is risk

perceptions and the influence of past returns with company stock. Employees significantly underestimate

36

the risks of their own company’s stock, and they are also unduly and erroneously influenced by past stock

performance. The findings regarding overconfidence and aversion to realizing losses may also come into

play with company stock. When workers have too rosy a view of the future but have trouble selling their

company stock at a loss, it is unlikely that providing additional information will quickly alter these

attitudes and produce changes in investment portfolios.

As in the auto-pilot 401(k) case, one policy option would be to provide employees with an

optional statutory mechanism that automatically reduces their exposure to company stock to a given

percentage that declined with age (e.g., 20 percent or 10 percent of assets by age 65). For example, a plan

might offer a provision that drew down the participant’s position steadily each quarter over some

prescribed period, say three or five years. In other words, participants may need a pre-commitment

device that works automatically for them as they near retirement, a concept dubbed as the “sell more

tomorrow” idea by the originators of the “save more tomorrow” plan.

4. Sensible plan design includes default choices at retirement. Current policy has permitted the

conversion of pensions from plans that pay life annuities, into programs that give workers a choice to

receive their life-long saving in a lump sum at retirement. Behavioral research suggests that annuity

versus lump sum decisions at retirement could be better framed, taking into account participants’

understanding of mortality versus investment risks. One question has to do with what should be the

default choice. In DB plans, the default has traditionally been an annuity, though more plans are now

offering lump-sum options. In DC plans, the default is generally a lump sum, with no annuity option. An

alternative approach might be to frame the default as some mixture of annuity and lump sum, rather than

as an either-or decision. Of course, it would be essential to ensure that the two options are compared on

an “apples-to-apples” basis to avoid framing bias. To better preserve the longevity protection that

pensions once offered, policymakers may find it sensible to make annuitization the default, and to make

loans against the pension accruals more difficult to obtain.

In sum, this research overview on behavioral decisionmaking finds important and valuable new

insights into how plan sponsors, benefit plan consultants, and policymakers must rethink pension plan

37

design in this new century. It seems clear that participants can be better served, when they make the hard

decisions to accumulate and decumulate retirement assets. It is because retirement saving decisions are at

least an order of magnitude more complex than other economic decisions, that people need help. The

thought process requires a sequence of critical savings and investment decisions over a lifetime, backed

by a complex and ongoing forecast of needs and resources. It should not be surprising, then, that for a

substantial segment of the workforce, this task proves daunting and discouraging. And for some people,

the problem may be so complex that they are forced to rely on decision heuristics that simplify decision-

making, but may not necessarily produce appropriate outcomes. While much has been learned, a central

question remains: how can the various stakeholders strike the best balance between encouragement and

compulsion? What system can both preserve participant decision-making while offering limits on choice,

so as to encourage the outcomes that rational and forward thinking consumers would want for

themselves? This volume offers much to inform the debate.

The authors are grateful for comments provided by Shlomo Benartzi. Opinions expressed are solely those of the authors. Financial support for this research was provided by the National Bureau of Economic Research and the Pension Research Council. This study is part of the NBER program on the Economics of Aging. Mitchell and Utkus, 2003. All Rights Reserved.

38

References Agnew, Julie Pierluigi Balduzzi, and Annika Sundén. 2003 “Portfolio Choice and Trading in a

Large 401(k) Plan.” American Economic Review 93(1): 193-215. Aizcorbe, Ana M., Arthur B. Kennickell, and Kevin B. Moore. 2003. “Recent Changes in U.S.

Family Finances: Evidence from the 1998 and 2001 Survey of Consumer Finances.” Federal Reserve Bulletin January: 1-32.

Ameriks, John. This volume. "How Do Retirees Go from Stock to Flow?" Ameriks, John, Andrew Caplin, and John Leahy. 2003. "Wealth Accumulation and the

Propensity to Plan." Quarterly Journal of Economics: 1007-1047. Arkes, Hal R., Neal V. Dawson, Theodore Speroff, Frank E. Harrell, Jr., Carlos Alzola, Russell

Phillips., Norman Desbiens, Robert K. Oye and Alfred F. Connors, Jr. 1995. “The Covariance Decomposition of the Probability Score and its Use in Evaluating Prognostic Estimates.” Medical Decision Making 15(2): 120-31.

Baber, Brad M. and Terrance Odean. 2000. “Trading is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors.” Journal of Finance LV(2): 773-806.

Banks, J., Robert Blundell, and S. Tanner. 1998. “Is There a Retirement-Savings Puzzle?” American Economic Review 88(4) September: 769-788.

Barber, Brad M. and Terrance Odean. 2001. “Boys will be Boys: Overconfidence and Common Stock Investment.” Quarterly Journal of Economics February: 261-292.

Barberis, Nick and Ming Huang. 2001."Mental Accounting, Loss Aversion, and Individual Stock Returns," Journal of Finance (56) 4: 1247-1292.

Barberis, Nicholas and Richard Thaler. 2002. “A Survey of Behavioral Finance.” NBER Working Paper 9222.

Benartzi, Shlomo. 2001. “Excessive Extrapolation and the Allocation of 401(k) Accounts to Company Stock.” Journal of Finance. LVI(5): 1747-1764.

Benartzi, Shlomo, and Richard Thaler. 2001. “Naive Diversification Strategies in Retirement Saving Plans,” American Economic Review, March 91(1): 79-98.

_____. 2002. "How Much Is Investor Autonomy Worth?" Journal of Finance (57) 4: 1593-1616.

_____. Forthcoming. ”Save More Tomorrow.” Journal of Political Economy. Bernheim, B. Douglas, Jonathan Skinner and Steven Weinberg. 2001. “What Accounts for the

Variation in Retirement Wealth among U.S. Households?” American Economic Review 91 (4): 832-857.

Bodie, Zvi. 1995. “On the Risk of Stocks in the Long Run.” Financial Analysts Journal (51)3: 18-22.

Bodie, Zvi, Brett Hammond, and Olivia S. Mitchell. 2002. “A Framework for Analyzing and Managing Retirement Risks”. In Innovations in Financing Retirement, eds. Zvi Bodie, Brett Hammond, Olivia S. Mitchell, and Stephen Zeldes. Philadelphia, PA: University of Pennsylvania Press: 3-19.

Brown, Jeffrey, Olivia S. Mitchell, and James Poterba. 2000. “The Role of Real Annuities and Indexed Bonds in an Individual Accounts Retirement Program.” In Risk Aspects of Investment-Based Social Security Reform, eds. John Y. Campbell and Martin Feldstein. NBER: University of Chicago Press: 321-360

39

_____. 2002. “Mortality Risk, Inflation Risk, and Annuity Products.” In Innovations in Financing Retirement, eds. Zvi Bodie, Brett Hammond, Olivia S. Mitchell and Stephen Zeldes. Philadelphia, PA: University of Pennsylvania Press: 175-197.

.Brown, Jeffrey, Olivia S. Mitchell, James Poterba, and Mark Warshawsky. 1999. “Taxing Retirement Income: Nonqualified Annuities and Distributions from Qualified Accounts”. National Tax Journal LII (3) September: 563-592.

_____. 2001. The Role of Annuity Markets in Financing Retirement. Cambridge, MA: MIT Press.

Campbell, John Y. and Luis M. Viceira. 2002. Strategic Asset Allocation: Portfolio Choice for Long-Term Investors. Oxford: Oxford University Press.

Choi, James, David Laibson, Brigitte Madrian and Andrew Metrick. 2001a. “Defined Contribution Pensions: Plan Rules, Participant Decisions, and the Path of Least Resistance.” NBER Working Paper 8655.

_____. 2001b. “For Better or For Worse: Default Effects and 401(k) Savings Behavior.” NBER Working Paper 8651.

_____. 2003. “Passive Decisions and Potent Defaults.” NBER Working Paper 9917. _____. This volume. “Employee Investment Decisions about Company Stock.” Clark, Robert L. Clark, Madeleine B. d’Ambrosio, Ann A. McDermed and Kshama Sawant.

This volume. “Sex Differences, Financial Education, and Retirement Goals.” Congressional Budget Office (CBO). 2003. “Utilization of Tax Incentives for Retirement

Saving.” August. www.cbo.gov. Davis, Steven J. and Paul Willen. 2002. “Income Shocks, Asset Returns, and Portfolio Choice.”

In Innovations in Retirement Financing, eds. Zvi Bodie, Brett Hammond, Olivia Mitchell and Stephen Zeldes. Philadelphia, PA: University of Pennsylvania Press: 20-49.

De Bondt, Werner. 1998. “A Portrait of the Individual Investor.” European Economic Review 42: 831-844.

Drinkwater, Matthew Drinkwater and Eric T. Sondergeld. This volume. “Perceptions of Mortality Risk: Implications for Annuities.”

Duflo and Saez. This volume. “Implications of Information and Social Interactions for Retirement Saving Decisions.”

Dus, Ivica, Raimond Maurer, and Olivia S. Mitchell. Forthcoming. “Betting on Death and Capital Markets in Retirement: A Shortfall Risk Analysis of Life Annuities versus Phased Withdrawal Plans.” Pension Research Council Working Paper, in preparation.

Employee Benefit Research Institute (EBRI). 2003. Retirement Confidence Survey: A Summary of Results. http://www.ebri.org/rcs/2003/03rcssof.pdf

Frederick, Shane, George Loewenstein and Ted O’Donoghue. 2002. “Time Discounting and Time Preference: A Critical Review.” Journal of Economic Literature. 40(2): 351-401.

Goetzmann, William N. and Alok Kumar. 2001. “Equity Portfolio Diversification.” NBER Working Papers 8686.

Huberman, Gur and Paul Sengmueller. 2003. “Company Stock in 401k Plans.” Columbia University Working Paper.

Hurd, Michael D. and Kathleen McGarry. 1995. “Evaluation of Subjective Probability Distributions in the HRS.” Journal of Human Resources. 30(0): S268-92.

Hurd, Michael and Susann Rohwedder. 2003. “The Retirement-Consumption Puzzle: Anticipated and Actual Deadlines in Spending at Retirement.” NBER Working Paper No. w9586.

40

Kahneman, Daniel. 2003. “The Psychology of Risky Choices.” Address before the Investment Company Institute, May 2003, Washington DC.

Kahneman, Daniel and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decision Under Risk.” Econometrica 47(2) March: 263-91.

_____. 2000. Choices, Values and Frames. Cambridge, Ma: Russell Sage Foundation and Cambridge University Press.

Laibson, David I. 1997. “Golden Eggs and Hyperbolic Discounting.” Quarterly Journal of Economics. 112 (2) May: 443-478.

Laibson, David I., Andrea Repetto, and Jeremy Tobacman. 1998. ”Self Control and Saving for Retirement.” Brookings Papers on Economic Activity I: 91-196.

Lichtenstein, Sarah, Baruch Fischhoff, and Lawrence Phillips. 1982. "Calibration of Probabilities: The State of the Art to 1980." In Judgment Under Uncertainty: Heuristics and Biases, eds. Daniel Kahnemann, Paul Slovic, and Amos Tversky. Cambridge: Cambridge University Press: 306-334.

Loewenstein, George and Drazen Prelec. 1992. “Anomalies in Intertemporal Choice: Evidence and an Interpretation.” Quarterly Journal of Economics 57: 573-98.

Lusardi, Annamaria . This volume. “Saving and the Effectiveness of Financial Education.” McGill, Daniel, Kyle Brown, John Haley and Sylvester Schieber. Forthcoming. Fundamentals of

Private Pensions, 8ed. Oxford: Oxford University Press. MacFarland, Donna, Carolyn D. Marconi, and Stephen P. Utkus. This volume. “’Money

Attitudes’ and Retirement Plan Design: One Size Does Not Fit All.” Madrian, Brigitte C. and Dennis F. Shea. 2001. “The Power of Suggestion: Inertia in 401(k)

Participation and Savings Behavior.” Quarterly Journal of Economics. 116(4): 1149-1187. MetLife. 2003. The MetLife Retirement Income IQ Test: Findings from the 2003 National Survey

of American Pre-Retirees. New York: Metlife Mature Market Institute, June. Mitchell, Olivia S. and David McCarthy. 2003 Forthcoming. “Annuities for an Ageing World.”

In Developing an Annuities Market in Europe, eds. E. Fornero and E. Luciano. London: Elgar.

Mitchell, Olivia S., James Moore, and John Phillips. 2000. “Explaining Retirement Saving Shortfalls”. In Forecasting Retirement Needs and Retirement Wealth, eds. Olivia S. Mitchell, Brett Hammond, and Anna Rappaport. Philadelphia, PA: Univ. of Pennsylvania Press: 139-166

Mitchell, Olivia S., James Poterba, Mark Warshawsky, Jeffrey Brown .1999. “New Evidence on the Money’s Worth of Individual Annuities.” American Economic Review. December: 1299-1318.

Mitchell, Olivia S. and Stephen P. Utkus. Forthcoming. “Company Stock and Retirement Plan Diversification.” In The Pension Challenge: Risk Transfers and Retirement Income Security, eds. Olivia S. Mitchell and Kent Smetters. Oxford: Oxford University Press.

Modigliani, Franco and Brumberg, R. 1954. “Utility Analysis and the Consumption Function: An Interpretation of Cross-section Data.” In Post-Keynesian Economics, ed. K.K. Kurihara New Brunswick, NJ: Rutgers University Press.

Moore, James and Olivia S. Mitchell. 2000. “Projected Retirement Wealth and Saving Adequacy”. In. Forecasting Retirement Needs and Retirement Wealth, eds. Olivia S. Mitchell, Brett Hammond and Anna Rappaport. Philadelphia, PA: Univ. of Pennsylvania Press: 68-94.

41

Moore, James H. Jr. and Leslie A. Muller. 2002. “An Analysis of Lump-sum Pension Distribution Recipients.” Monthly Labor Review May: 29-46.

Mullainathan, Sendhil and Richard H. Thaler. 2000. “Behavioral Economics.” NBER Working Paper 7948.

Odean, Terrance. 1998a. “Are Investors Reluctant to Realize Their Losses?” Journal of Finance 53(5): 1775-1798.

_____. 1998b. ”Volume, Volatility, Price and Profit: When All Traders are Above Average.” Journal of Finance 53(5): 1887-1933.

Panis, Stan. This volume. “Annuities and Retirement Satisfaction” Patel, Jayendu, Richard Zeckhauser, and Darryll Hendricks. 1991. “The Rationality Struggle:

Illustrations from Financial Markets.” American Economic Review 81: 232-236. Polkovnichenko, Valery. 2003. "Household Portfolio Diversification." Working paper

presented at the Rodney White Center for Financial Research conference, Household Portfolio Choice and Financial Decision-Making, March.

Poterba, James, Joshua Rauh, Steven Venti, and David Wise. 2003. “Utility Evaluation of Risk in Retirement Saving Accounts.” NBER Working Paper 9892.

Purcell, Patrick J. 2002. “The Enron Bankruptcy and Employer Stock in Retirement Plans.” CRS Report for Congress. US GOP. Code RS21115.

Saliterman, Victor and Barry Sheckley. This volume. “Adult Learning Principles and Pension Participant Behavior.”

Samuelson, Paul A. 1989. “The Judgment of Economic Science on Rational Portfolio Management: Timing and Long-Horizon Effects.” Journal of Portfolio Management 16: 4-12.

Scott, Jason and Greg Stein. This volume. “The Impact of Advice on Employee Behavior and Retirement Prospects.”

Selnow, Gary. This volume. “Motivating Retirement Planning: Problems and Solutions.” Sethi-Iyengar, Sheena, Gur Huberman, and Wei Jiang. This volume. “How Much Choice is Too

Much? Contributions to 401(k) Retirement Plans.” Shefrin, Hersh. 2003. Beyond Greed and Fear: Understanding Behavioral Finance and the

Psychology of Investing. Oxford: Oxford University Press. Shefrin Hersh and Meir Statman. 1985. “The Disposition to Sell Winners Too Early and Ride

Losers Too Long: Theory and Evidence.” Journal of Finance 40 (3) July: 777-790. Simon, Herbert A. 1955. “A Behavioral Model of Rational Choice.” Quarterly Journal of

Economics 69 February: 99-118. Sondergeld, Eric and Matthew Drinkwater. This volume. “Perceptions of Mortality Risk and

Implications for Annuities” Statman, Meir. This volume. “. Behavioral Portfolios: Hope for Riches and Protection from

Poverty.” Svenson, Ernest O. 1981. “Are We All Less Risky and More Skillful Than Our Fellow

Drivers?” Acta Psychologica 47: 143-148. Taylor, Shelley E. and Brown, Jonathan D. 1988. “Illusion of Well-Being: A Social

Psychological Perspective on Mental Health.” Psychological Bulletin 103: 193-210. Thaler, Richard H. 1981. “Some Empirical Evidence on Dynamic Inconsistency.” Economic

Letters 8: 201-207. Reprinted in Richard Thaler, Quasi-Rational Economics. New York: Russell Sage Foundation: 1994.

42

Thaler, Richard H. and Shlomo Benartzi. 1999. "Risk Aversion or Myopia? Choices in Repeated Gambles and Retirement Investments." Management Science 45(3): 364-381.

_____. Forthcoming. “Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving.” Journal of Political Economy.

Thaler, Richard and Eric Johnson. 1990. “Gambling with the House Money and Trying to Break Even: The Effects of Prior Outcomes on Risky Choices.” Management Science 36(6): 643-660. Reprinted in Richard Thaler, Quasi-Rational Economics. New York: Russell Sage Foundation: 1994.

Thaler Richard and H.M. Shefrin. 1981. “An Economic Theory of Self-Control.” Journal of Political Economy 89: 392-406.

Thaler, Richard and Cass R. Sunstein. 2003. “Libertarian Paternalism.” American Economic Review 93(2): 175-179.

Tiger, Lionel. 1979. Optimism: The Biology of Hope. New York. Tversky, Amos and Daniel Kahneman. 1974. “Judgment Under Uncertainty: Heuristics and

Biases.” Science 185: 1124-1131. Warner, John T. and Saul Pleeter. 2001. “The Personal Discount Rate: Evidence from Military

Downsizing Programs.” American Economic Review 91(1): 33-53. Weber, Elke. This volume. “Who’s Afraid of a Poor Old Age? Risk Perception and Risk

Management Decisions.” Weinstein, N. 1980. “Unrealistic Optimism About Future Life Events.” Journal of Personality

and Social Psychology 39: 806-820.

43

$0.00

$1.00

$2.00

$3.00

$4.00

$5.00

$6.00

0 5 10 15 20 25 30

Years

Exponential discounter

Hyperbolic discounter

0% time value of money

$5.74

$3.32

Figure 1. Exponential versus Hyperbolic Discounters: Growth of $1 Over Time

Source: Authors’ calculations.

37%

86%

0%

20%

40%

60%

80%

100%

Before automatic enrollment After automatic enrollment

Part

icip

atio

n R

ate

Figure 2. Decision Framing: The Impact of Automatic Enrollment on New Hire Plan Participation Rates Source: Madrian and Shea (2002)

44

3.5%

6.5%

9.4%

11.6%

13.6%

0%

4%

8%

12%

16%

Initial 1st increase 2nd increase 3rd increase 4th increase

Figure 3. The Impact of Commitment Devices and Inertia: The Impact of SmarT on Plan Saving Rates Source: Thaler and Benartzi (forthcoming)

65% 67% 68% 66%71% 69% 69% 70% 72% 67%

58%48%

0%

25%

50%

75%

100%

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

6/20

03

Figure 4. Anchoring and Adjustment: Current Equity Contributions By Plan Entry Date (% contribution allocated to equity investments )

Source: Vanguard Group, 2003.

45

Value

GainsLosses

Figure 5. Prospect Theory Value Function Source: Authors’ calculations.

46

Table 1. The Self-Control Problem: Divergence between Desired and Actual Behavior

Action Planned Change (%) Actual Change (%)

Enroll in 401(k) plan 100 14

Increase contribution rate 28 8

Change fund selection 47 15

Change fund allocation 36 10

Source: Choi et. al. (2001a), Table 6.

47

Table 2. Perceptions of Company Stock Risk and Return Q. Would you say your employer’s stock is more risky, less risky or has about the same level of risk as an investment in a diversified stock fund with many different stocks? (n=415)

Participant Report: Level of Risk in Company Stock**

% of Participants

Actual Av. St. Dev. of

Company Stock (%)*

Actual Av. Company

Stock Return(%)*

More risky 33 40 ** -8.8 **Same level of risk 42 36 ** -2.0 **Less risky 22 31 ** 2.2 **Don't know 3 35 -6.0 TOTAL 100% S&P 500 18 ** -1.1 ** * Returns and standard deviations of participants’ company stock returns for the five-year period ending Sept. 30, 2003. Standard deviation calculated over 60 months and annualized. ** “More risk,” “same level of risk” and “less risky” categories are all significantly different from one another at the 95 or 99 percent level. Standard deviations are all significantly higher than the S&P 500 at the 99 percent level. Returns for “more risky”(“less risky”) are significantly lower (higher) than the S&P 500 at the 99 percent level. Source: Vanguard Group (2003).

48

Endnotes 1 For a recent review see Kahneman and Tversky (2000); Mullainathan and Thaler (2000);

Shefrin (2003) and Barberis and Thaler (2002). For a review of retirement and portfolio

implications, see also Statman (this volume), Duflo and Saez (this volume), and Choi et. al. (this

volume).

2 Indeed the 2002 Nobel Prize in economics awarded to Daniel Kahneman and Vernon Smith,

recognized the far-reading importance of this shift in paradigm. The Prize Committee noted

“[t]raditionally, much of economic research has relied on the assumption of a “homo

œconomicus” motivated by self-interest and capable of rational decision-making. Economics has

also been widely considered a non-experimental science, relying on observation of real-world

economies rather than controlled laboratory experiments. Nowadays, however, a growing body

of research is devoted to modifying and testing basic economic assumptions... This research has

its roots in two distinct, but currently converging, areas: the analysis of human judgment and

decision-making by cognitive psychologists, and the empirical testing of predictions from

economic theory by experimental economists.” (http://www.mea.uni-

mannheim.de/winter/lehre/03-ss/behav.htm)

3 Aizcorbe et al. (2003) summarize results from the 2001 Survey of Consumer Finances, and he

finds for example, age-related fluctuations in net worth, financial assets and debt consistent with

the life-cycle hypothesis.

4 There are methodological or substantive caveats regarding this research, yet on the whole the

research does indicate that at least some households are not particularly good at solving the

retirement saving problem. Hurd and Rohwedder (2003) find, for example, that households

anticipate a 20% drop in consumption, and so the decline in retiree consumption may be

49

rationally anticipated, not unexpected. One concern about Bernheim etal. (2001) is that it uses

data on food consumption (both in and out of the home) as a proxy for total consumption.

5 See also Loewenstein and Prelec (1992), Laibson (1997), and Frederick, Loewenstein and

O’Donoghue (2002). As Thaler recalls, doubts among economists about the consistency of

individual time preferences dates back at least to Robert Strotz, who in the mid 1950s wrote that

“special attention should be given…to a discount function…[that] ‘overvalues’ the more

proximate satisfaction relative to the more distant ones…..” (cited in Thaler, 1981, p. 127).

6 Two examples are drawn from insurance and health care (Thaler and Sunstein, 2003). In the

US, in the state of New Jersey, the default option under the state’s car insurance regulations

offers workers a limited right to sue for damages (with a lower insurance premium). In the

neighboring state of Pennsylvania, the default is the regular right to sue; to obtain the limited

right to sue and the lower premium, the car driver must make a positive election. In New Jersey,

20 percent of individuals retain the full right to sue, which requires a positive election, while in

Pennsylvania, 75 percent retain the full right to sue, which is the default. Similarly, the donation

of vital organs upon one’s death is common in a number of European countries, with over 90

percent of individuals donating organs; yet it is comparatively rare in the US, with less than 20

percent making organ donations. In the European countries, organ donation is the default; in the

US, it requires a positive election.

7 Goetzman and Kumar, 2001; De Bondt, 1998.

8 See Thaler and Johnson (1990) for a discussing of break-even thinking and the house money

effect.

9 See Arkes et al. (1995), Odean (1998a), Svenson (1981), Tiger (1979) and Weinstein (1980).

Taylor and Brown (1988) point out that over-optimism offers psychological benefits as well.

50

10 For an overview of the overconfidence literature, particularly on the ability of individuals to

calibrate their forecasts, see Lichenstein, Fischoff and Phillips (1982). Odean (1998a, 1998b)

also summarizes findings on overconfidence.

11 Shefrin (2003) summarizes both these conclusions from De Bondt (1998) and related literature

on overconfidence in investing.

12 It may be worth noting that opting for a lump sum permits retirees to avoid using the unisex

mortality tables required by law for employee benefit plans. This could be an appealing result for

those who anticipate living less long than the combined male/female life expectancy. Taking the

lump sum and spending it also makes the near-poor retiree more likely to be eligible for social

welfare benefits payable to the indigent.